On Wednesday 2008 December 24 16:16:26 ow...@netptc.net wrote: > AFAIK the MTU is set by the initiator of the IP connection and any > intermediate router can't change it. > Larry
Not exactly. The MTU is really a link-level property that's immutable. However, path MTU discovery is not required for IPv4, so the MTU is generally just set to a reasonable value at the endpoint. However, any intermediate (or end-point) node can fragment any packet. Intermediate routers tend to do that anytime the packet size exceeds the link MTU that they choose to route down. The alternative is to down the packet and, IIRC, send back and ICMP message that indicates the packet exceeds the MTU (which, in IPv4 is generally ignored or lost). In IPv6, path MTU discovery is required and fragmentation is only allowed at the end points. Intermediate routers are required to drop (or re-route) packets that exceed the MTU, and send a MTU notification (over ICMPv6) back. The end-point is required to use this packet to adjust it's path MTU to allow packets to be send via the smaller link. IPv6 is not implemented widely yet, and there are certainly issues that may continue to prevent it from being so. We need (or could certainly use) the extra addresses, but there's a lot more to that involved in IPv6 -- AAAA records in DNS, mandated IPSec (I think), aforementioned constraints on intermediate nodes, etc. -- Boyd Stephen Smith Jr. ,= ,-_-. =. b...@iguanasuicide.net ((_/)o o(\_)) ICQ: 514984 YM/AIM: DaTwinkDaddy `-'(. .)`-' http://iguanasuicide.net/ \_/
signature.asc
Description: This is a digitally signed message part.
